bug#15155: 24.3; wrap-prefix in adaptive-wrap-prefix-mode with variable-pitch has wrong face

[Frank Fischer]

Am Fri, 23 Aug 2013 09:41:01 +0300
schrieb Eli Zaretskii <eliz@gnu.org>:

> > Date: Thu, 22 Aug 2013 22:32:46 +0200
> > From: Frank Fischer <frank.fischer@mathematik.tu-chemnitz.de>
> > Cc: monnier@iro.umontreal.ca, 15155@debbugs.gnu.org
> > 
> > Am Thu, 22 Aug 2013 19:56:05 +0300
> > schrieb Eli Zaretskii <eliz@gnu.org>:
> > 
> > > Can you build Emacs from the trunk?  I would like to ask you to
> > > try a patch that might fix this in your use case.
> > 
> > Sure, I'd be happy to help ;)
> 
> Thank you.  Please try the patch below.  It should solve the problem
> with the font; you will still need to make sure on the Lisp level that
> the prefix has the right face, to get the color you want.

That patch seems to work, thank you.

I have only one (very) minor issue left: depending on the mode the
wrap-prefix may contain characters with different faces (e.g. in
emacs-lisp mode !!!;;; would be a correct prefix, the first three
characters with default face, the latter three with comment-face), hence
there is not really "the one" correct face to assign 'wrap-prefix with
`propertize`. That why I tried the variant

(setq adaptive-fill-function
  (lambda ()
    (and adaptive-fill-regexp
         (looking-at adaptive-fill-regexp)
         (match-string 0))))

hoping that (match-string 0) returns the whole prefix with the correct
fontification. However, there are situations where this does not work:
for example if one has a (very long) wrapped line without comment and
then inserts a semicolon (only one) at the beginning of that line to
comment it out, then the wrap-prefix does not get comment face (to be
precise, it has the correct font, thus your patch works, but not the
correct color). But if one inserts a second semicolon then the
wrap-prefix gets the correct color. I assume this happens because the
wrap-prefix is set before font-lock changed the face of inserted
semicolon. 

Perhaps this is a more an issue of adaptive-wrap-mode, I
don't know. If my guess is correct, is it possible to run the
appropriate hooks (I think adaptive-wrap uses
`jit-lock-register`) *after* font-lock changed the face? Or is there
another, better way to detect the correct face?

If you don't have a solution to this question, don't worry. I'm quite
happy with your patch and it does already what I want ;)

Thank you again,
Frank